This is the implementation of tsGCN proposed in our paper:

Shiping Wang, Zhihao Wu, Yuhong Chen, and Yong Chen*, Beyond Graph Convolutional Network: An Interpretable Regularizer-Centered Optimization Framework, AAAI 2023.

Full paper with appendix can be found HERE.

• Python == 3.9.12
• PyTorch == 1.11.0
• Numpy == 1.21.5
• Scikit-learn == 1.1.0
• Scipy == 1.8.0
• Texttable == 1.6.4
• Tensorly == 0.7.0
• Tqdm == 4.64.0

python main.py

• --device: number of gpus or 'cpu'.
• --path: path of datasets.
• --dataset: name of datasets.
• --seed: random seed.
• --fix_seed: fix the seed or not.
• --n_repeated: number of repeated times.
• --model: choose the model, GCN or tsGCN.
• --bias: enable bias.
• --lr: learning rate.
• --weight_decay: weight decay.
• --num_pc: number of labeled samples per class.
• --num_epoch: number of training epochs.

All the configs are set as default, so you only need to set --dataset and --model. For example:

python main.py --dataset Cora --model tsGCN

• ACM
• BlogCatalog
• Citeseer
• Cora
• CoraFull
• Flickr
• Pubmed
• UAI

Please unzip the datasets folders first.

Saved in ./datasets/datasets.7z

@inproceedings{
    wu2023tsGCN,
    title={Beyond Graph Convolutional Network: An Interpretable Regularizer-Centered Optimization Framework},
    author={Shiping Wang, Zhihao Wu, Yuhong Chen, Yong Chen},
    booktitle={Proceedings of the 37th AAAI Conference on Artificial Intelligence},
    year={2023},
}